Re: [PATCH] trace-cmd: Add support for built-in plugins

From: Steven Rostedt <rostedt@goodmis.org>
To: vincent.donnefort@arm.com
Cc: linux-trace-devel@vger.kernel.org
Subject: Re: [PATCH] trace-cmd: Add support for built-in plugins
Date: Mon, 11 Nov 2019 18:02:24 -0500	[thread overview]
Message-ID: <20191111180224.2992867b@gandalf.local.home> (raw)
In-Reply-To: <1573126532-373130-1-git-send-email-vincent.donnefort@arm.com>

Hi Vincent,

On Thu,  7 Nov 2019 11:35:32 +0000
vincent.donnefort@arm.com wrote:

> From: Vincent Donnefort <vincent.donnefort@arm.com>
> 
> When building trace-cmd with the -static option, plugins might not work, if
> the shared version for the Glibc is not available on the system. To still
> give a chance to have plugins, a new built-in mechanism is introduced.
> 
> The built-in build support for a plugin can be easily enabled with the
> declaration:
> 
>  TEP_PLUGIN(plugin_name)
>  {
>          plugin_init()
>  }
> 
>  /* optional */
>  TEP_PLUGIN(plugin_name)
>  {
>          plugin_exit()
>  }
> 
> Enabling as a first step, blk and sched_switch plugins as built-in.

I don't have a problem with the idea of this patch. My concern is that
the code in lib/traceevent will soon be obsoleted by using the
libtraceevent library in the Linux kernel under tools/lib/traceevent.

That means when we transition to the libtraceevent library, this work
will be lost.

Instead, could this be added to tools/lib/traceevent in the Linux
kernel somehow, and instead of building it into the executable, perhaps
make it part of the libtraceevent.a file (the static version of the
libary).

Thoughts?

-- Steve